Evaluation and Validation of Semi-Supervised Ant-inspired Sentence-Level Sentiment Prediction Clustering
Mohammed Qasem, Parimala Thulasiraman · 2019
Exact algorithmic clustering approaches are not affordable for many real-world applications, requiring innovative, approximation methods. Among them evolutionary techniques and semi-supervised learning approaches have led to improved performance on several real world applications. In this paper we combine these two approaches to design a semi-supervised clustering algorithm to predict sentiments in sentence-level product reviews. We evaluate and validate the technique to sentence-level sentiment analysis problem and show that compared to baseline techniques, multi-class logistic regression and lexicon based approaches, our technique outperforms by 20%.